During the last three decades of coral reefs studies, the large areal coverage of data derived from satellite images has increasingly been used to complement the more detailed but spatially limited data produced by conventional fieldwork. Continuous improvement in sensor capabilities, along with the development of increasingly refined methods for image processing, has lead to ever more accurate maps of physical and biological variables of importance to reef ecology. During the same period, an abundance of field studies have documented statistical relationships between aspects of the reef habitat and its fish community. Despite numerous stochastic influences, such as spatially concentrated and temporally variable fish recruitment pulses or the selective and patchy mortality caused by fishing, several aspects of habitat have been shown to significantly influence the fish community. Fortunately the most important of these, water depth, the structural complexity of the reef, and the cover of live coral, are possible to estimate from currently available satellite imagery. The research presented in the following pages has combined the statistical relationships between the fish community and its habitat with the capability of satellite imagery to map that habitat, thereby answering the research question: How can remote sensing be used to map coral reef fish communities? In the process, a set of new techniques for predictive modeling of complex relationships have been compared, the influence of a range of habitat variables on the fish community quantified, the spatial scales at which the fish-habitat relationships are strongest have been explored, and new methods for deriving estimates of some aspects of the coral reef habitat from satellite imagery have been developed. The results presented in this thesis thus contribute to the further understanding of fish-habitat relationships, while providing a template for producing spatially explicit predictive models of fish community variables. This is not only of scientific interest, but also of substantial value to the conservation community that tries to protect the world’s remaining healthy coral reef ecosystems, and their fish communities, from an array of man-made influences.

Heterogeneity, the degree of dissimilarity, is one of the most important and widely applicable concepts in ecology. It is highly related to ecosystem conditions and features wildlife habitat. Grasslands have been described as inherently heterogeneous because their composition and productivity are highly variable across multiple scales. Therefore, biological heterogeneity can be an indicator of ecosystem health. The mixed prairie in Canada, characterized by its semiarid environment, sparse canopy, and plant litter, offers a challenging region for environmental research using remote sensing techniques. This thesis dwells with the plant canopy heterogeneity of the mixed prairie ecosystem in the Grasslands National Park (GNP) and surrounding pastures by combining field biological parameters (e.g., grass cover, leaf area index, and biomass), field collected hyperspectral data, and hierarchical resolution satellite imagery. The thesis scrutinized four aspects of heterogeneity study: the importance of scale in grassland research, relationships between biological parameters and remotely collected data, methodology of measuring biological heterogeneity, and the influence of climatic variation on grasslands biological heterogeneity. First, the importance of scale is examined by applying the semivariogram analysis on field collected hyperspectral and biophysical data. Results indicate that 15 - 20 m should be the appropriate resolution when variations of biological parameters and canopy reflectance are sampled. Therefore, it is reasonable to use RADARSAT-1, Landsat TM, and SPOT images, whose resolutions are around 20 m, to assess the variation of biological heterogeneity. Second, the efficiency of vegetation indices derived from SPOT 4 and Landsat 5 TM images in monitoring the northern mixed prairie health was examined using Pearsons correlation and stepwise regression analyses. Results show that the spectral curve of the grass canopy is similar to that of the bare soil with lower reflectance at each band. Therefore, vegetation indices are not necessarily better than reflectance at green and red wavelength regions in extracting biological information. Two new indices, combining reflectance from red and mid infrared wavelength regions, are proposed to measure biological parameters in the northern mixed prairie. Third, texture analysis was applied to quantify the biological variation in the grasslands. The textural parameters of RADARSAT imagery correlated highly with standard deviation of the field collected canopy parameters. Therefore, textural parameters can be applied to study the variations within the mixed prairie. Finally, the impacts of climatic variation on grassland heterogeneity at a long time scale were evaluated using Advanced Very High Resolution Radiometer (AVHRR) , Normalized Difference Vegetation Index (NDVI), Maximum Value Composite (MVC), and SPOT Vegetation NDVI MVC imagery from 1993 to 2004. A drought index based on precipitation data was used to represent soil moisture for the study area. It was found that changes of temperature and precipitation explain about 50% of the variation in AVHRR NDVI (i.e., temporal heterogeneity) of the northern mixed prairie. Trend line analysis indicates that the removal of grazing cattle carry multiple influences such as decreasing NDVI in some parts of the upland and valley grassland and increasing NDVI in the valley grassland. Results from this thesis are relevant for park management by adjusting grassland management strategies and monitoring the changes in community sizes. The other output of the thesis is furthering the remote sensing investigation of the mixed prairie based on information of the most appropriate resolution imagery.

Part of the problem with implementing a rangeland insurance program is that the acreage of different pasture types, which is required in order to determine an indemnity payment, is difficult to measure on the ground over large areas. Remote sensing techniques provide a potential solution to this problem. This study applied single-date SPOT (Satellite Pour IObservation de la Terre) imagery, field collected data, and geographic information system (GIS) data to study the classification of land cover and vegetation at species level. Two topographic correction models, Minnaert model and C-correction, and two classifying algorithms, maximum likelihood classifier (MLC) and artificial neural network (ANN), were evaluated. The feasibility of discriminating invasive crested wheatgrass from natives was investigated, and an exponential normalized difference vegetation index (ExpNDMI) was developed to increase the separability between crested wheatgrass and natives. Spectral separability index (SSI) was used to select proper bands and vegetation indices for classification. The results show that topographic corrections can be effective to reduce intra-class rediometric variation caused by topographic effect in the study area and improve the classification. An overall accuracy of 90.5% was obtained by MLC using Minnaert model corrected reflectance, and MLC obtained higher classification accuracy (~5%) than back-propagation based ANN. Topographic correction can reduce intra-class variation and improve classification accuracy at about 4% comparing to the original reflectance. The crested wheatgrass was over-estimated in this study, and the result indicated that single-date SPOT 5 image could not classify crested wheatgrass with satisfactory accuracy. However, the proposed ExpNDMI can reduce intra-class variation and enlarge inter-class variation, further, improve the ability to discriminate invasive crested wheatgrass from natives at 4% of overall accuracy. This study revealed that single-date SPOT image may perform an effective classification on land cover, and will provide a useful tool to update the land cover information in order to implement a rangeland insurance program.

Habitat loss and human-caused mortality are the most serious threats facing grizzly bear (<i>Ursus arctosi</i> L.) populations in Alberta, with conflicts between people and bears in agricultural areas being especially important. For this reason, information is needed about grizzly bears in agricultural areas. The objectives of this research were to find the best possible classification approach for determining multiple classes of agricultural and herbaceous land cover for the purpose of grizzly bear habitat mapping, and to determine what, if any, spatial and compositional components of the landscape affected the bears in these agricultural areas. Spectral and environmental data for five different land-cover types of interest were acquired in late July, 2007, from Landsat Thematic Mapper satellite imagery and field data collection in two study areas in Alberta. Three different classification methods were analyzed, the best method being the Supervised Sequential Masking (SSM) technique, which gave an overall accuracy of 88% and a Kappa Index of Agreement (KIA) of 83%. The SSM classification was then expanded to cover 6 more Landsat scenes, and combined with bear GPS location data. Analysis of this data revealed that bears in agricultural areas were found in grasses / forage crops 77% of the time, with small grains and bare soil / fallow fields making up the rest of the visited land-cover. Locational data for 8 bears were examined in an area southwest of Calgary, Alberta. The 4494 km2 study area was divided into 107 sub-landscapes of 42 km2. Five-meter spatial resolution IRS panchromatic imagery was used to classify the area and derive compositional and configurational metrics for each sub-landscape. It was found that the amount of agricultural land did not explain grizzly bear use; however, secondary effects of agriculture on landscape configuration did. High patch density and variation in distances between neighboring similar patch types were seen as the most significant metrics in the abundance models; higher variation in patch shape, greater contiguity between patches, and lower average distances between neighboring similar patches were the most consistently significant predictors in the bear presence / absence models. Grizzly bears appeared to prefer areas that were structurally correlated to natural areas, and avoided areas that were structurally correlated to agricultural areas. Grizzly bear presence could be predicted in a particular sub-landscape with 87% accuracy using a logistic regression model. Between 30% and 35% of the grizzlies‟ landscape scale habitat selection was explained.

In 1998, hairy prairie-clover was listed as threatened by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC) and subsequently afforded protection under the Species at Risk Act in 2004. Hairy prairie-clover, being a habitat specialist species confined to areas of sparsely vegetated to bare sand, may provide an indication of the loss of a once viable natural mixed-grass prairie and sand dune landscape indicative to southern Saskatchewan. Therefore, critical habitat identification for hairy prairie-clover is of particular concern, not only to provide conservation efforts for this particular species, but also for bare sand and sand dune environments which are some of the most sensitive landscapes on the prairies. The goal of this thesis is to identify and spatially delineate areas of critical habitat for hairy prairie-clover within the range of a known metapopulation in the Dundurn sand hills south of Saskatoon, Saskatchewan. This research was divided into two specific objectives: 1) to investigate the spatial relationship between bare sand habitat for hairy prairie-clover and other land cover classes, and 2) to study the relationship between habitat configuration and hairy prairie-clover occurrence.<p> To achieve the first objective, the desired output was a land cover classification of the study site at an appropriate spatial and temporal resolution. Wavelet analysis revealed that the optimum spatial resolution for bare sand identification and delineation in the study site was between 2-5 m. Analysis of field spectroradiometer measurements throughout the growing season concluded that the early and late growing seasons were best for spectrally discriminating between land cover classes. A multi-resolution, multi-temporal land cover classification using object-oriented methods resulted in an overall classification accuracy of 79% with a users and producers accuracy of 85% for bare sand. Grassland comprised the matrix of the area covering 45.5% of the study site. Aspen and shrub were the most dominating landscape elements comprising 25.5% and 19.2% of the study site respectively. Bare sand made up only 6.0% of the study site while juniper was the least persistent class comprising only 2.7% of the study site.<p> The desired output from objective two was a critical habitat landscape mosaic for hairy prairie-clover. Patch scaled metrics were calculated for bare sand patches identified in the land cover classification from objective one. Binary logistic regression was used to identify which metrics could explain and predict hairy prairie-clover occurrences. Results showed that almost 29% of the variation in bare sand patch occupancy could be explained by the size, shape, and degree of isolation of a sand patch as well as the amount of vegetation on a sand patch in the early growing season. Based on these variables, 18.8% of sand patches in the study site were predicted to be unsuitable hairy prairie-clover habitat, 45.7% were predicted to be marginally unsuitable, 32.7% were predicted to be suitable, and 2.8% were predicted to be marginally suitable. Overall prediction accuracy was about 61% with 80% of occurrences and 54% of non-occurrences being correctly predicted.

Phosphorus flowing via rivers into the Baltic Sea is a major source of nutrients, and in some cases the limiting factor for the growth of algae which causes the phenomenon known as eutrophication. Remote sensing of phosphorus, here using Landsat TM-data, can help to give a better understanding of the process of eutrophication. Since Landsat TM-data is used, this could form a basis for further spatio-temporal analysis in the Baltic Sea region. A method originally described and previously applied for a Chinese river is here transferred and applied to three different rivers flowing into the Baltic Sea. The results show that by measuring the proxy variables of Secchi Depth and Chloryphyll-a the remote sensing model is able to explain 41% of the variance in total- phosphorus for the rivers Dalälven, Norrström and Gavleån without any consideration taken to CDOM, turbidity or other local features.

A special challenge posed by mineral dust aerosols is associated with their predominantly nonspherical particle shapes. In the present study, the scattering and radiative properties for nonspherical mineral dust aerosols at violet-to-blue (0.412, 0.441, and 0.470 μm) and red (0.650 μm) wavelengths are investigated. To account for the effect of particle nonsphericity on the optical properties of dust aerosols, the particle shapes for these particles are assumed to be spheroids. A combination of the T-matrix method and an improved geometric optics method is applied to the computation of the single-scattering properties of spheroidal particles with size parameters ranging from the Rayleigh to geometric optics regimes. For comparison, the Mie theory is employed to compute the optical properties of spherical dust particles that have the same volumes as their nonspherical counterparts. The differences between the phase functions of spheroidal and spherical particles lead to quite different lookup tables (LUTs) involved in retrieving dust aerosol properties. Moreover, the applicability of a hybrid approach based on the spheroid model for the phase function and the sphere model for the other phase matrix elements is demonstrated. The present sensitivity study, employing the Moderate Resolution Imaging Spectroradiometer (MODIS) observations and the fundamental principle of the Deep Blue algorithm, illustrates that neglecting the nonsphericity of dust particles leads to an underestimate of retrieved aerosol optical depth at most scattering angles, and an overestimate is noted in some cases. The sensitivity study of the effect of thin cirrus clouds on dust optical depth retrievals is also investigated and quantified from MODIS observations. The importance of identifying thin cirrus clouds in dust optical depth retrievals is demonstrated. This has been undertaken through the comparison of retrieved dust optical depths by using two different LUTs. One is for the dust only atmosphere, and the other is for the atmosphere with overlapping mineral dust and thin cirrus clouds. For simplicity, the optical depth and bulk scattering properties of thin cirrus clouds are prescribed a priori. Under heavy dusty conditions, the errors in the retrieved dust optical depths due to the effect of thin cirrus are comparable to the assumed optical depth of thin cirrus clouds. With the spheroidal and spherical particle shape assumptions for mineral dust aerosols, the effect of particle shapes on dust radiative forcing calculations is estimated based on Fu-Liou radiative transfer model. The effect of particle shapes on dust radiative forcing is illustrated in the following two aspects. First, the effect of particle shapes on the single-scattering properties of dust aerosols and associated dust direct radiative forcing is assessed, without considering the effect on dust optical depth retrievals. Second, the effect of particle shapes on dust direct radiative forcing is further discussed by including the effect of particle nonsphericity on dust optical depth retrievals.

The majority of biofuels are produced from corn and grain. The drawback to these sources of biofuels is the vast amount of cultivated land needed to produce substantial amounts of biofuel, potentially increasing the price of food and livestock products. Mesquite trees, a type of woody plant, are a proven source of bioenergy feedstock found on semi-arid lands. The overall objectives of this study were to develop algorithms for determining woody plant biomass on rangelands in Texas at plot-level using terrestrial lidar and at the local scale by integrating reference biomass and multispectral imagery. Terrestrial lidar offers a more efficient method for estimating biomass than traditional field measurements. Variables from the terrestrial lidar point cloud were compared to ground measurements of biomass to find a best fitting regression model. Two processing methods were investigated for analyzing the lidar point cloud data, namely: 1) percentile height statistics and 2) a height bin approach. Regression models were developed for variables obtained through each processing technique for estimating woody plant, above-ground biomass. Regression models were able to explain 81 percent and 77 percent of the variance associated with the aboveground biomass using percentile height statistics and height bins, respectively. The aboveground biomass map was generated by using the cokriging interpolation method with NDVI and ground biomass data. According to cross-validation, ordinary cokriging estimated biomass accurately (R^2 = 0.99). The results of this study revealed that terrestrial lidar can be used to accurately and efficiently estimate the aboveground biomass of mesquite trees in a semi-arid environment at plot level. Moreover, spatial interpolation techniques proved useful in scaling up biomass estimates to local scale.

Over the past century, glaciers throughout the tropics have predominately retreated. These small glaciers, which respond quickly to climate changes, are becoming increasingly important in understanding glacier-climate interactions. The glaciers on Mt. Jaya in Irian Jaya, Indonesia are the last remaining tropical glaciers in the Western Pacific region. Although considerable research exists investigating the climatic factors most affecting tropical glacier mass balance, extensive research on the Mt. Jaya glaciers has been lacking since the early 1970s. Using IKONOS satellite images, the ice extents of the Mt. Jaya glaciers in 2000, 2002, 2003, 2004, and 2005 were mapped. The mapping indicates that the recessional trend which began in the mid-19th century has continued. Between 1972 (Allison, 1974; Allison and Peterson, 1976) and 2000, the glaciers lost approximately 67.6% of their area, representing a reduction in surface ice area from 7.2 km2 to 2.35 km2. From 2000 to 2005, the glaciers lost an additional 0.54 km2, representing approximately 24% of the 2000 area. Rates of ice loss, calculated from area measurements for the Mt. Jaya glaciers in 1942, 1972, 1987, and 2005, indicate that ice loss on Mt. Jaya has increased during each subsequent period. Preliminary modeling, using 600 hPa atmospheric temperature, specific humidity, wind speeds, surface precipitation, and radiation values, acquired from the NCEP Reanalysis dataset, indicates that the only climate variable having a statistically-significant change with a magnitude great enough to strongly affect ice loss on these glaciers was an increase in the mean monthly atmospheric temperature of 0.24°C between 1972 and 1987. However, accelerated ice loss occurring from 1988-2005 without large observed changes in the weather variables indicates that a more complex explanation may be required. Small, though statistically-significant changes were found in regional precipitation, with precipitation decreasing from 1972-1987 and increasing from 1988-2005. While, individually, these changes were not of sufficient magnitude to have greatly affected ice loss on these glaciers, increased precipitation along with a rising freezing level may have resulted in a greater proportion of the glacier surface being affected by rain. This may account for the increased recession rate observed in the latter period.

Thesis (M.S.)--Mississippi State University. Department of Electrical and Computer Engineering. / Title from title screen. Includes bibliographical references.